Antibody humanisation: a case of the 'Emperor's new clothes'? by Mike Clark

Transcription

1 ViewPoint / Review article for August 2000 special issue of Immunology Today This revised version submitted 1st June 2000 Antibody humanisation: a case of the 'Emperor's new clothes'? by Mike Clark Keywords antibody engineering; humanisation; human antibodies; chimeric antibodies; anti-idiotype; HAMA; HAHA; V-regions Summary The antiglobulin response is perceived as a major problem in clinical development of therapeutic antibodies. Successive technical developments such as chimeric, humanised and now fully human antibodies claim to offer improved solutions to this problem. Although there is clear evidence that chimeric antibodies are less immunogenic than murine monoclonal antibodies, little evidence exists to support claims for further improvements as a result of more elaborate humanisation protocols. Mike Clark, 'The Emporer's New Clothes' - 1 -

2 In the fable of the 'Emperor's New Clothes', a tailor tricks the Emperor into believing that he has a new and magical set of clothes, when in fact he does not. Most of the Emperor's subjects are equally taken in by the charade, too gullible, or too frightened to question what they are told. One small boy says what he actually sees and the trickery is eventually exposed. Without taking the analogy too far, I argue that antibody engineering has provided therapeutic antibodies with many new sets of clothes, and there are many competing claims as to which are better than the others. Not all these claims have a sound scientific basis and there is little evidence of rigorous scientific proof in support of some of these claims, particularly regarding immunogenicity of antibody variable regions. It is 25 years since Kohler and Milstein published their work on the use of cell fusion for the production of monoclonal antibodies from immunised mice 1. The technique was rapidly and widely adopted and has provided an enormous repertoire of useful research reagents [Reviewed in this issue by Little et al]. In turn, these reagents have formed a key element in the rapid progress of our understanding of biology. Although murine-derived monoclonal antibodies have been widely applied in clinical diagnostics, they have had limited success in human therapy [Reviewed in this issue by Glennie and Johnson]. Two major problems in the therapeutic use of murine monoclonal antibodies were identified early on. First, although murine antibodies had exquisite specificity for therapeutic targets, they did not always trigger the appropriate human effector systems of complement and Fc receptors 2. Second, even when murine antibodies were identified that did work in vivo, the patient's immune system normally cut short the therapeutic window since murine antibodies were recognised by a human anti-murine-antibody immune response (HAMA) (refs 3,4). An obvious exception to these generalisations has been the success of the mouse monoclonal antibody Orthoclone OKT-3 used in prevention of organ graft rejection and which, in 1986, was the first monoclonal antibody to be approved by the FDA for clinical use [see review by Glennie and Johnson, this issue]. However, in parallel with the advances in the production of monoclonal antibodies from hybridomas, other major technological advances were happening in recombinant DNA technology. More was being understood of how immunoglobulin genes were organised and expressed by B-cells and about how germline immunoglobulin genes were rearranged, and mutated, to form the repertoire of functional immunoglobulin genes 5,6. Inevitably the techniques of monoclonal antibody production and recombinant DNA technology were combined to try to resolve the problems that had arisen in the application of murine antibodies to human therapy 7-9. Thus, the field of antibody engineering was initiated and today many of the products currently in clinical development by the biotechnology industry are recombinant monoclonal antibodies, or their derivatives [see article in this issue by Glennie and Johnson]. It seems clear that recombinant DNA technology as applied to antibody constant (C-) regions has provided solutions to the problems of antibody functions in-vivo 10,11. But was this the right way to go for antibody variable (V-) regions, or is it a case of the "Emperor's new clothes"? Did enthusiasm to find ever more elaborate solutions to the HAMA response blind scientists to a more obvious shortcoming in the whole approach - namely human anti-human antibody responses (HAHA), in particular to the idiotype 12? Also is an anti-idiotype response only a function of the V-region sequence of the antibody or is immunogenicity more likely to depend on other factors 12,13? IgG is the preferred class for therapeutic antibodies for several practical reasons 11. IgG antibodies are very stable, easily purified, and stored. In-vivo they have a long biological half-life which is not just a function of their size but is also a result of their interaction with the so-called Brambell receptor (or FcRn) (refs 14, 15). This receptor seems to protect IgG from catabolism within cells and recycles it back to the plasma. In addition, IgG has subclasses which are able to interact with and trigger a whole range of humoral and cellular Mike Clark, 'The Emporer's New Clothes' - 2 -

3 effector mechanisms. These mechanisms are initiated through immune complex formation, activation of complement, and also through binding to cellular Fc receptors and complement receptors 11. Figure 1 shows the basic structural features of a human IgG1. Figure 2 is a cartoon illustrating how recombinant DNA technology can be used to transform a murine antibody into a human antibody for therapeutic applications. The immunoglobulin molecule is made up of a set of globular domains. Thus antibody engineering can be applied to the immunoglobulin molecule in a modular fashion to generate a chimeric antibody with murine V-regions and human C-regions. Humanised antibodies can be generated where the antigenbinding complimentarity determining regions (CDRs) are murine, while the rest of the antibody including the V-region framework regions (FRs) is human. Figure 1 Model of an IgG molecule A spacefilling model of an IgG molecule coloured according to the structural features. Constant regions are in blue, carbohydrate in purple, V-region frameworks are in red and complimentarity determining regions in yellow. Mike Clark, 'The Emporer's New Clothes' - 3 -

4 Figure 2 Schematic drawing of humanisation of an IgG molecule The blue colour represents mouse sequences and the red colour human sequences. In a chimeric antibody the mouse heavy and light chain V-region sequences are joined onto human heavy chain and light chain C-regions. In humanised antibodies the mouse CDR sequences (3 from the heavy chain V- region and 3 from the light chain V-region) are grafted onto human V-region FRs and expressed along with human C-regions. Chimeric antibodies with human effector functions The overall functions of an antibody are related to the constant regions which determine the class and sub-class. Clearly, human effector functions have evolved alongside human immunoglobulins. Very early during the development of rodent monoclonal antibodies for human therapy, it became clear that most monoclonal antibodies were ineffective in situations where cell destruction was a desired outcome (eg in cancer therapy). In some situations where several similar antibodies were used to target the same antigen, cell depletion was dependent on the antibody class and subclass 16. Several studies have shown that human IgG1 is the preferred choice for chimeric antibodies in situations where activation of effector functions is the desired outcome There may be other occasions in therapy when the effector activity of IgG1 is not required. In this situation it is possible to make use of the human subclasses IgG2 and IgG4, which lack some of the effector functions 11. However, all four human IgG subclasses mediate at least some biological functions. Antibody engineering has again been applied to this problem and antibodies with altered Fc regions, and thus altered activity, have been produced. For example, removal of the conserved N-linked glycosylation site on the constant region of a CD3 antibody reduces the unwanted biological side effects associated with FcR binding, whilst retaining efficacy in blocking T-cell function 17. An alternative strategy has recently been described whereby potent blocking antibodies can be generated by making antibodies with sequences derived from a combination of IgG1, IgG2 and IgG4 18. Humanisation to avoid the immune response Immunogenicity of therapeutic antibodies is a significant problem and severely limits their widespread and repeated application to treat many diseases 12,13. Factors likely to be important in immunogenicity are listed in Box 1. Many believe that the immunogenicity of an immunoglobulin is mainly a function of its foreignness to the human patient (top three items in Box 1). Most strategies for engineering antibodies to reduce immunogenicity are based upon the immunological concept that we are tolerant of all self proteins and only respond to foreign proteins. Most immunologists would, I hope, recognise that this basic idea Mike Clark, 'The Emporer's New Clothes' - 4 -

5 is flawed, particularly for immunoglobulins. By definition, every clone of B-cells with a unique specificity has a unique V-region sequence, it possesses a unique 'idiotype'. New clones of B-cells are generated every day throughout life by the processes of somatic gene recombination and somatic mutation 5,6. It seems implausible that tolerance to each B-cell clone is generated as soon as every new sequence emerges. Indeed, an equally familiar immunological concept is the opposite extreme; that every clone of B-cells provokes an antiidiotype response, which in turn provokes another anti-idiotype, thus forming an antibody network which thus regulates immune responses 19. The true situation is probably somewhere between these two extremes. Box 1. Factors likely to influence immunogenicity of therapeutic antibodies * Murine constant regions V-region sequences Human Ig allotypes Unusual glycosylation Method of administration Frequency of administration Dosage of antibody Patients' disease status Patients' immune status Patients' MHC haplotype Specificity of antibody Cell surface or soluble antigen? Formation of immune complexes with antigen Complement activation by antibody Fc receptor binding by antibody Inflammation and cytokine release * All of these factors are likely to have a bearing on the immunogenicity of a therapeutic antibody. Many of them may be an inevitable consequence of the treatment, and the desired mode of action of the antibody. Only the first three on the list are altered by humanisation and the fourth is likely to depend on the expression system used for antibody production. Four basic antibody engineering strategies have been adopted to tackle the immunogenicity of therapeutic antibodies. In chimeric antibodies, the murine constant regions are replaced with human constant regions, on the basis that the constant region contributes a significant component to the immunogenicity 7,8,12. As a further step on from chimeric antibodies, V- region humanisation, or reshaping, involves changing some of the FR residues of the V- regions to sequences considered more 'human' whilst retaining the CDR sequences necessary for antigen binding [see Figure 2.] (refs 9,20-22). Two additional strategies provide what some call 'fully human V-regions' - human antibody V-regions can be selected from phage libraries by affinity selection on antigen 23,24 ; transgenic mice have been constructed which have had their own immunoglobulin genes replaced with human immunoglobulin genes so that they produce human antibodies upon immunisation 25,26. Mike Clark, 'The Emporer's New Clothes' - 5 -

6 Immunogenicity versus antigenicity of therapeutic antibodies Before discussing just how sensible it is to promulgate the idea that 'V-region humanisation' or 'fully human V-regions' are the way ahead we need to consider the immunological difference between the concepts of 'antigenicity' and 'immunogenicity'. Antigenicity is defined in terms of the ability of a molecule to be recognised by preformed antibodies. A given antigen may be recognised by many different antibody specificities and these define the antigenic 'epitopes'. Immunogenicity refers to the ability of an antigen to provoke an immune reponse. It is possible for an antigen to be highly immunogenic, weakly immunogenic, non-immunogenic or tolerogenic 27. For example, high doses of deaggregated human IgG given intravenously to mice can be tolerogenic but the same antigen given in aggregated form subcutaneously can be highly immunogenic. It is the immunogenicity which is important for therapy, not the antigenicity 12,13,27. For protein antigens such as immunoglobulins the immune response is usually T-cell dependent and there are many factors which might have a significant influence on the immunogenicity including the need for appropriate antigen processing and presentation and 'secondary signals'. How might recombinant antibodies influence these requirements? We shouldn't lose sight of the fact that antibodies have evolved to play a key role in the immune response to infection and thus they have features which often enhance immunogenicity of antigens bound up in immune complexes with the antibodies. It seems unlikely that at the same time as enhancing immunogenicity of the bound antigen that the antibody is completely inert with regard its own immunogenicity (particularly for its idiotype). Binding to Fc receptors 28, and the activation of the complement cascade 29,30, are two obvious ways that antibodies can enhance immune responses. Also, antibodies which directly bind to cell surfaces may be internalised and processed for presentation 27 whilst antibodies that cause cell killing and destruction may give rise to what is sometimes referred to as 'Danger' signals 31. In support of these ideas it has been reported that the presence of the murine Fc region may make antibodies more immunogenic and also that Fab antibodies are much less immunogenic than are Fab' 2 or whole antibodies 12. Also antibodies which have been engineered so as not to activate effector functions are less immunogenic 32. Unfortunately, those properties of antibodies which result in enhanced antigen presentation are likely to be related to the desired properties of the antibodies which have been deliberately chosen for their therapeutic effect 9,10,16. It may thus prove very difficult to generate tolerance even to humanised or fully human antibodies which cause such effects in-vivo. There is some hope from strategies whereby tolerance can be induced to a non-cell binding, non-inflammatory antibody which has a slightly modified but otherwise similar idiotype to a cell binding therapeutic antibody 13,27,33. However the therapeutic antibody will still have a unique idiotype even compared to the deliberately engineered tolerising antibody and so at best this strategy might reduce or delay the probability of an anti-idiotype response but is unlikely to eliminate it in all therapeutic circumstances. Other strategies involve identifying and engineering out the major peptide sequences which would activate helper T-cells when presented on major histocompatibility complex (MHC) Class II molecules. The problem here is that there are so many different alleles of MHC Class II in the human population that it would only be possible to do this for a limited number of the more common epitopes. Humanised versus chimeric antibodies So are 'humanised' antibodies, or even so called 'fully human' antibodies, likely to be better than 'chimeric' antibodies for human therapy? This has implications at several levels. Humanisation of the V-regions may be considered to try to reduce potential problems of immunogenicity 9,20. However, during the humanisation process the antibody affinity is Mike Clark, 'The Emporer's New Clothes' - 6 -

7 frequently reduced 9. This reduction in affinity may be minimised by careful selection of human FRs that are homologous to the starting antibody 21,22,34 or by reintroduction of important murine FR residues back into the engineered antibody 9,21,34. This strategy may result in a humanised antibody which is itself very homologous to a chimeric form of the same antibody. This can be easily seen in Table 1 where the murine V-regions of the antibodies anti-tac and OKT3 have a similar overall percentage homology to a closest match human germline gene (75% and 72%) compared with the homology of the humanised antibodies and the acceptor myeloma protein sequences (72% and 69%). Indeed I am almost tempted to argue that humanisation of these two antibodies has reduced their homology. The humanisation methods used have maximised the homology for FRs at the expense of the homology for CDRs, however the overall percentage homologies for chimeric and humanised versions of these antibodies are similar. Furthermore, the homology of chimeric antibodies compared to the human germline is only slightly less than some human myeloma protein sequences compared to the germline. This is presumably a consequence of somatic mutations in the myeloma protein sequences. Table 1. Percentage homology* of humanised and chimeric antibody V-regions Antibody comparisons FR CDR Complete V-region 1. Humanised V H (i.e. CDR grafted) versus human myeloma V H Campath-1H versus NEW 98% 9% 74% Anti-Tac versus EU 86% 28% 72% OKT3 versus KOL 85% 25% 69% 2. Chimeric V H (i.e. murine) versus human germline V H Campath-1G 78% 41% 68% Anti-Tac 77% 69% 75% OKT3 77% 56% 72% 3. Human myeloma V H versus human germline V H NEW 87% 50% 78% EU 95% 77% 91% KOL 91% 81% 88% *The percentage amino acid homology for several antibody heavy chain V- regions are shown (from Routledge et al 1993) (ref 34). 1. Humanised antibodies (Campath-1H, anti-tac, OKT3) shown compared to the human myeloma protein sequences used as acceptor sequences for the humanisation (NEW, EU, KOL). 2. Murine sequences compared to the closest matching human germline immunologlobulin sequences found in the databases, i.e. the homology for chimeric versions of the antibodies. 3. Comparison of the three human myeloma sequences against human germline immunoglobulin sequences. Mike Clark, 'The Emporer's New Clothes' - 7 -

8 The similar homology of humanised and chimeric forms of an antibody as shown in Table1 arises because V-regions are often much more homologous between species than are C- regions. Although this seems counter-intuitive the explanation is simple, there are many different human V-regions to compare with any one chosen murine V-region, but there are only four human IgG C-regions. However, there is little scientific evidence to support a substantial reduction in immunogenicity when the penalty is an antibody of perhaps lower affinity 9,21,22 and one where the commercial viability is reduced further by the need to pay out numerous royalties for licenses which cover the antibody engineering process. Humanisation may increase the production costs of a therapeutic antibody substantially. If these cost increases are passed on to patients, some healthcare providers might consider the treatment too expensive to be routinely used. Then, ironically, although the percentage of patients showing antiglobulin responses may be reduced, the total numbers of patients treated will also be reduced. My recommendation is that at present it is well worth doing a costbenefit analysis for humanisation on every new antibody and if humanisation means that the costs outway the benefit, then it may well be better to produce and use a chimeric form of the antibody for therapy. Testing the efficacy of V-region humanisation What would constitute a 'proper scientific test' of humanisation of V-regions as a means of reducing the immunogenicity? At a minimum, this would probably require the generation of humanised and chimeric forms of several different antibodies with different idiotypes. These pairs of antibodies would then have to be used in similar clinical studies where the number of patients is high enough to establish reliable immunogenicity frequencies. A significant result would indicate that all the humanised antibodies are less immunogenic than their chimeric counterparts. Such an experiment is unlikely to be contemplated on commercial or on ethical grounds. Furthermore, there is already considerable evidence for a wide variability in measured immunogenicity of chimeric and humanised antibodies to different antigens and used in different diseases 12,13,38. Thus it can already be concluded that other factors listed in Box 1 may be more important than the V-region sequences. So what can be done? We can return to animal experiments. It is possible to make antiidiotype antibodies to rodent antibodies by immunising the same strains from which the original monoclonal antibodies were derived 39. Thus rodents are clearly not tolerant of antibodies from the same species and strain. However, deliberately generating an antiidiotype is often not easy, and usually requires the use of adjuvant or coupling to 'carrier' proteins such as keyhole limpet haemocyanin (KLH) (ref 39). In contrast it is also possible in rodents to generate tolerance to foreign proteins, including human immunoglobulins, as mentioned earlier 13,27,33. Animal models suggest that the nature of the target antigen is more important than the sequence of the antibody V-regions, with cell binding antibodies being the most difficult with which to achieve tolerance 27. Alternative strategies for producing 'human' antibodies Other strategies for the production of ' fully human' antibodies, include phage libraries 23,24, or transgenic mice 25,26, both making use of repertoires of human V-regions. A third technique consists of reconstituting severe combined immunodeficient (SCID) mice with human peripheral blood lymphocytes, immunising these animals with antigen, followed by rescuing the immune B-cells by fusion with myeloma cells 40. However, antibodies generated using these methods still have unique idiotypes, and in the case of antibodies from transgenic mice (and of course antibodies from reconstituted SCID mice), the somatic mutations are in FRs as well as in the CDRs (ref 26). Some like to claim that chimeric antibodies are overall about 75% human in sequence (% homology of whole IgG), whilst humanised antibodies Mike Clark, 'The Emporer's New Clothes' - 8 -

9 are 95% human and antibodies from transgenic mice or phage display are 100% human. These figures can only be derived if murine and human antibodies are thought of as being totally different (Figure 2). In fact, as illustrated in Table 1, there is considerable sequence homology between mouse and human immunoglobulin sequences, as would be obvious to anyone who looks at the sequence databases 22,34. The Kabat database shows clearly that many V-region sequences, particularly for FRs, are conserved between species 41. This, combined with the large repertoire and diversity of sequences in each species, makes it likely that most rodent sequences have a homologous human counterpart (see Table 1) ( refs 22,34-37,41). Equally, somatic mutations in antibodies from transgenic and reconstituted mice, particularly during affinity maturation, means that they are no longer 100% identical to inherited human germline genes, as is the case with human myeloma proteins (see Table 1) (ref 34). The other problem with use of human antibodies generated by these methods, is that it might still be necessary to engineer them as chimeric constructs to provide them with a suitable human C-region for therapy, particularly where novel C-regions are contemplated 11,17,18. Again this has cost implications as it might then require royalty license payments to cover both technologies. Concluding remarks For some therapeutic antibodies it seems likely that the problems of immunogenicity are likely to remain whatever the strategies chosen for their production. However the antiidiotype response is generally only a problem where repeated treatments are required for chronic and relapsing diseases such as in therapy of autoimmune disease 12,13. Indeed for many therapeutic applications in acute disease situations, the possibility of an anti-idiotype response is not likely to have a major impact on their efficacy in the vast majority of patients 13,38. I would not want to imply that there is an absolute correlation between the fable of the Emperor's New Clothes and antibody engineering, but I think the scientific question of how the immunogenicity of an antibody V-region relates to its sequence must be addressed. It is unlikely, I argue, to be a simple matter of percentage homology. It is interesting to reflect on a News and Views article in Nature, written by Alan Munro to accompany the publication of the first descriptions of recombinant 'chimeric' antibodies 42. Although he used the term 'chimaeric' where today we would refer to 'humanised', with some foresight, he wrote, 'by using the techniques of genetic engineering, it may be possible to obtain antibodies in which the antigen -binding site is defined by sequences from a rodent monoclonal antibody of the right specificity whereas the rest of the molecule is as 'human' in structure as possible.... Possibly the human immune system will not recognize such chimaeric molecules as foreign, but there are good reasons to think that they will'. Yet what was obvious to Alan Munro in 1984 seems to have been forgotten or is ignored by many today. Acknowledgements I thank my colleagues over the years for many stimulating discussions on antibody therapy and immunological tolerance. More importantly, I acknowledge the contributions of the undergraduates to whom I have taught immunology. Their probing questions bring home the shortcomings that still remain in our understanding of immunological problems. Affiliation Mike Clark (mrc7@cam.ac.uk, is in the Immunology Division, Department of Pathology, Cambridge University, Tennis Court Road, Cambridge CB2 1QP. Mike Clark, 'The Emporer's New Clothes' - 9 -

10 References 1 Kohler, G. and Milstein, C. (1975) Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256, Clark M. et al (1984) Advantages of rat monoclonal antibodies. Immunol. Today 4, Schroff, R.W. et al (1985) Human anti-murine responses in patients receiving monoclonal antibody therapy. Cancer Res 45, Shawler, D.L. et al (1985) Human immune responses to multiple injections of murine monoclonal immunoglobulins. J. Immunol. 135, Tonegawa, S. (1983) Somatic generation of antibody diversity. Nature 302, Alt, F. W. et al (1992) VDJ recombination. Immunol. Today 13, Boulianne, G.L et al (1984) Production of functional chimaeric mouse/human antibody. Nature 312, Morrison, S.L. et al (1984) Chimeric human antibody molecules: Mouse antigenbinding domains with human constant region domains Proc. Natl. Acad. Sci. USA 81, Riechmann, L. et al (1988) Reshaping human antibodies for therapy. Nature 332, Hale, G. et al (1988) Remission induction in non-hodgkin lymphoma with reshaped human monoclonal antibody CAMPATH-1H. The Lancet 2, Clark, M.R. (1997) IgG Effector Mechanisms. Chemical Immunology 65, Kuus-Reichel, K. et al (1994) Will immunogenicity limit the use, efficacy, and future development of therapeutic monoclonal antibodies? Clin. Diagn. Lab. Immunol. 1, Isaacs J.D. (1990) The antiglobulin response to therapeutic antibodies. Sem. Immunol. 2, Ghetie, V. et al (1996) Abnormally short serum half-lives of IgG in beta-2- microglobulin-deficient mice. Eur. J. Immuol. 26, Junghans, R.P. and Anderson, C.L. (1996) The protection receptor for IgG catabolism is the beta2-microglobulin-containing neonatal intestinal transport receptor. Proc. Natl. Acad. Sci. USA 93, Dyer, M.J.S. et al (1989) Effects of CAMPATH-1 antibodies in vivo in patients with lymphoid malignancies: influence of antibody isotype. Blood 73, Bolt, S. et al (1993) The generation of a humanised, non-mitogenic CD3 monoclonal antibody which retains in vitro immunosuppressive properties. Eur. J. Immunol. 23, Armour, K.L. et al. (1999), Recombinant human IgG molecules lacking Fc-gamma Receptor I binding and monocyte triggering activities. Eur J Immunol 29, Macias, A. et al (1999) Novel cross-reactive anti-idiotype antibodies with properties close to human intravenous immunoglobulin (IVIg). Hybridoma 18, Jones, P.T. et al (1986) Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 321, Queen, C. et al (1989) A humanized antibody that binds to the interleukin 2 receptor. Proc. Natl. Acad. Sci. USA 86, Gorman, S.D.et al. (1991) Reshaping a therapeutic CD4 antibody. Proc. Natl. Acad. Sci. USA 88, Winter, G. et al (1994) Making antibodies by phage display technology. Annu Rev. Immunol. 12, Low, N.M. et al (1996) Mimicking somatic hypermutation: affinity maturation of Mike Clark, 'The Emporer's New Clothes'

11 antibodies displayed on bacteriophage using a bacterial mutator strain. J. Mol. Biol. 260, Bruggemann, M. et al (1991) Human antibody expression in transgenic mice: expression from 100kb of the human Igh locus. Eur J. Immunol. 21, Mendez, M.J. et al (1997) Functional transplant of megabase human immunoglobulin loci recapitulates human antibody response in mice. Nature Genetics 15, Benjamin, R.J. et al (1986) Tolerance to rat monoclonal antibodies: implications for serotherapy. J. Exp. Med. 163, van-vugt, M.J. et al (1999) The FcgammaRIa (CD64) ligand binding chain triggers major histocompatibility complex class II antigen presentation independently of it's associated FcR gamma-chain. Blood 94, Fearon, D.T. (1998) Non-structural determinants of immunogenicity and the B-cell co-receptors, CD19, CD21 and CD22. Adv. Exp. Med. Biol. 452, Lou, D. and Kohler, H. (1998) Enhanced molecular mimicry of CEA using photaffinity crosslinked C3d peptide. Nat Biotechnol. 16, Matzinger, P. (1994) Tolerance, danger, and the extended family. Annu. Rev. Immunol 12, Routledge, E.G. et al (1995) The effect of aglycosylation on the immunogenicity of a humanized therapeutic CD3 monoclonal antibody. Transplantation 60, Gilliland, L.K. et al (1999) Elimination of the immunogenicity of therapeutic antibodies. J. Immunol. 162, Routledge, E.G. et al (1993) Reshaping antibodies for therapy in: Protein Engineering of Antibody Molecules for Prophylactic and Therapeutic Applications in Man (Clark, M. ed), pp 13-44, Academic Titles, Nottingham UK. 35 Cook, G.P and Tomlinson, I.M. (1995) The human immunoglobulin VH repertoire. Immunol. Today 16, Williams, S.C. et al (1996) Sequence and evolution of the human germline V lambda repertoire. J. Mol. Biol. 264, Tomlinson, I.M. et al (1995) The structural repertoire of the human V kappa domain. EMBO J. 14, Rebello, P.R. et al (1999) Anti-globulin responses to rat and humanized CAMPATH- 1 monoclonal antibody used to treat transplant rejection. Transplantation 68, Cobbold, S.P. et al (1990) A simple method for measuring patient antiglobulin responses against isotypic or idiotypic determinants. Meth. Enzymol. 127, Lubin, I.H. et al (1994) Engraftment of human peripheral blood lymphocytes in normal strains of mice. Blood 83, Johnson, G. and Wu, T.T. (2000) Kabat Database and its applications: 30 years after the first variability plot. Nucleic Acids Research, 28, Munro, A. (1984) Uses of chimaeric antibodies. Nature 312, 597 Mike Clark, 'The Emporer's New Clothes'